Back to skill

Security audit

OpenClaw技能分类与统计

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed OpenClaw skill manager for inventory reports, with opt-in sharing and enable/disable actions that users should run deliberately.

Install only if you want a Bash-capable skill manager that can inspect your OpenClaw skills, write local reports/cache files, and, when explicitly requested, change skill enablement or upload a skill inventory to Feishu/Notion. Avoid --push in sensitive environments and review --enable/--disable targets before running them.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Rogue AgentSelf-Modification, Session Persistence
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (43)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The declared description describes a broad inventory and reporting tool for OpenClaw skills. The supplied code chunk does not implement or exercise those capabilities; it is a unit test focused on translator helper functions. Its observable behavior is limited to testing Chinese translation of descriptions and extraction of untranslated words. That is materially different from the declared primary purpose, so this is a clear mismatch.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'file_read' capability that is not listed in its permissions. This may indicate deceptive intent or missing permission declarations.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
88% confidence
Finding

This file performs configuration writes and backup creation, but the finding indicates those write capabilities are not declared in the skill's permissions model. Undeclared write capability is dangerous because users or the host agent may grant trust based on incomplete permissions metadata, while the skill can still modify persistent configuration and change enabled/disabled state of other skills.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
91% confidence
Finding

This file performs outbound network access via 'git ls-remote' to remote repositories, but the static finding indicates that this capability is not declared in the skill's permissions. Undeclared network access weakens user consent and platform trust boundaries, especially in a skill manager context that processes many installed skills and their metadata.

Content

No source excerpt is available for this finding.

Lp1

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'env' capability that is not listed in its permissions. This may indicate deceptive intent or missing permission declarations.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 30)May include surrounding context.

md
- 🌍 **Description Sinicization**: built-in glossary replaces common English phrases, with an untranslated-words report
- 📊 **Usage stats**: dual data source — `skill_usage` table / curator (`--usage`)
- 🔗 **Dependency graph**: missing-dependency edges as mermaid (md) and SVG bipartite chart (html) (`--deps`)
- 🔁 **Batch enable/disable**: writes `skills.entries.<name>.enabled` with auto-backup and post-write verification
- ⬆️ **Update check**: git track vs remote HEAD, ClawHub track hint — report only, never auto-upgrades (`--check-updates`)
- 📤 **Remote push**: opt-in push of exported Markdown to Feishu Docs / Notion pages (`--push`; prints a privacy notice before sending)

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 170)May include surrounding context.

Custom category overlay

Create ~/.openclaw/skill_categories.json (takes precedence over the built-in map):

json
{

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This markdown file contains natural-language guidance that fixes the skill's interface language to zh-CN as a design choice. The policy requires flagging language or locale constraints unless the skill offers a choice or clearly justifies a region-specific requirement, which is not shown here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest-style metadata sets language: zh-CN, which signals a fixed locale. Because the file does not offer an alternative locale or user opt-in around this restriction, it matches the language-policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The markdown explicitly states that output and category labels are in Chinese and targets Chinese users. Under the policy rule, forcing a specific language without offering user choice or opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The module docstrings and user-facing natural-language descriptions are entirely in Chinese, including the stated skill purpose and return semantics. This imposes a specific language/locale in the skill's natural-language interface without any opt-in, alternative language support, or documented region-specific justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

This code file contains natural-language docstrings and comments exclusively in Chinese, including the module description and function documentation. Under the policy, forcing a specific language without user opt-in or documented justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · data/openclaw.py (reported line 48)May include surrounding context.

python
for attempt in range(max_retries):
        try:
            result = subprocess.run(
                ['openclaw', 'skills', 'list', '--json'],
                capture_output=True,
                text=True,

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This function writes modified JSON back to the main OpenClaw configuration file, which can directly change skill enablement state. Although the module docstring and logger mention backup and write activity, there is no user-facing confirmation prompt or explicit disclosure at the point of the potentially impactful overwrite.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · data/toggle.py (reported line 84)May include surrounding context.

python
def subprocess_run_list_json():
    import subprocess
    out = subprocess.run(['openclaw', 'skills', 'list', '--json'],
                         capture_output=True, text=True, encoding='utf-8', timeout=120)
    if out.returncode != 0:
        raise RuntimeError(out.stderr.strip()[:200])

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The top-level docstring is written as an instruction-like description entirely in Chinese, which imposes a specific language/locale without any opt-in or alternative. Under the stated policy, forcing a specific language without user choice is a natural-language policy violation.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · data/updates.py (reported line 40)May include surrounding context.

python
if not os.path.isdir(os.path.join(skill_dir, '.git')):
        return ''
    try:
        out = subprocess.run(['git', '-C', skill_dir, 'rev-parse', 'HEAD'],
                             capture_output=True, text=True, timeout=30)
        if out.returncode == 0:
            return out.stdout.strip()

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
82% confidence
Finding

The code performs a network-capable git operation against a URL loaded from local metadata, which can be attacker-influenced if a malicious skill writes a crafted source-origin.json. Although _safe_git_url blocks obvious localhost/private IPv4 targets and requires HTTPS, it does not robustly prevent all SSRF-style destinations (for example via DNS resolution tricks, IPv6/private-address edge cases, or access to attacker-controlled hosts), so the skill may initiate unintended outbound connections during analysis.

Content

Scanner excerpt · data/updates.py (reported line 52)May include surrounding context.

python
def _remote_head(url: str) -> str:
    """远端默认分支 HEAD,失败返回空串"""
    try:
        out = subprocess.run(['git', 'ls-remote', url, 'HEAD'],
                             capture_output=True, text=True, timeout=60)
        if out.returncode == 0 and out.stdout:
            return out.stdout.split()[0]

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

This file’s natural-language content, including the module description and later user-facing log messages, is entirely in Chinese, with no indication that the skill supports other languages or that Chinese is a required locale. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · data/usage.py (reported line 58)May include surrounding context.

python
def _from_curator() -> Optional[Dict[str, Dict]]:
    """回退路:解析 curator status --json(字段名做防御式兼容)"""
    try:
        out = subprocess.run(
            ['openclaw', 'skills', 'curator', 'status', '--json'],
            capture_output=True, text=True, encoding='utf-8', timeout=60)
    except Exception as e:

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · exporter/notion.py (reported line 15)May include surrounding context.

python
from utils.webclient import http_json

logger = logging.getLogger(__name__)
API = 'https://api.notion.com/v1'
NOTION_VERSION = '2022-06-28'
MAX_BATCH = 100
RICH_TEXT_LIMIT = 2000

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The function reads a local Markdown file and sends its contents to the external Notion API, but the operational flow provides no confirmation prompt or user-facing disclosure at the point of transmission. While the module docstring describes the purpose generally, this function performs network transfer of user data silently once invoked.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The module docstring and CLI help text describe the tool as providing Chinese-localized reporting and untranslated-English extraction as built-in behavior, with no indication that users can choose another language or locale. This is a natural-language locale policy concern because the skill presents a fixed language preference rather than offering a language choice.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This test asserts that generated markdown must contain Chinese labels ('来源' and '调用量'), which semantically enforces a specific language in output. Under the policy, forcing a locale without user opt-in is a natural-language policy violation unless the locale constraint is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The assertion for '缺失依赖关系' indicates the generated dependency section is expected to use Chinese text. This suggests the skill's output is locked to a specific language, which violates the language-choice policy when no opt-in or justification is present in the file.

Content

No source excerpt is available for this finding.